have the FETs been replaced ?
how old is the power supply ?

The psu surely works fine (the whole system doesn't use more than 80w in full load anyway), Fet have not been replaced.
There are 2: F 9824 ND86030L. I couldn't find the datasheet, any advice about what I could use to change them?
Anyway so far the system apparently works fine, it crashes only when I use Winstone 99, but somtimes it slow down to a 486 level.

Nemo1985 wrote on 2021-04-21, 11:30:

The psu surely works fine (the whole system doesn't use more than 80w in full load anyway), Fet have not been replaced.
There are 2: F 9824 ND86030L. I couldn't find the datasheet, any advice about what I could use to change them?
Anyway so far the system apparently works fine, it crashes only when I use Winstone 99, but somtimes it slow down to a 486 level.

seems like it's a B, not an 8.
datasheet is here https://datasheetspdf.com/pdf-file/473679/Fai … hild/NDB6030L/1
52 A, 30 V. RDS(ON) = 0.0135 Ω @ VGS=10 V RDS(ON) = 0.020 Ω @ VGS=4.5

weedeewee wrote on 2021-04-21, 11:59:

seems like it's a B, not an 8. datasheet is here https://datasheetspdf.com/pdf-file/473679/Fai … hild/NDB6030L/1 52 A, 30 V. RD […]
Show full quote

Nemo1985 wrote on 2021-04-21, 11:30:

The psu surely works fine (the whole system doesn't use more than 80w in full load anyway), Fet have not been replaced.
There are 2: F 9824 ND86030L. I couldn't find the datasheet, any advice about what I could use to change them?
Anyway so far the system apparently works fine, it crashes only when I use Winstone 99, but somtimes it slow down to a 486 level.

seems like it's a B, not an 8.
datasheet is here https://datasheetspdf.com/pdf-file/473679/Fai … hild/NDB6030L/1
52 A, 30 V. RDS(ON) = 0.0135 Ω @ VGS=10 V RDS(ON) = 0.020 Ω @ VGS=4.5

Thank you for the precious tip!
The only one similar I could find is the IRL3803PBF (https://www.tme.eu/Document/c5e4ca68b7beda6d5 … 00a/irl3803.pdf) do you think it would be suitable?
I touched the mosfet and they aren't even warm, i'm actually using a cyrix cpu so it should stress the components (v2.9)

Nemo1985 wrote on 2021-04-21, 13:08:

The only one similar I could find is the IRL3803PBF (https://www.tme.eu/Document/c5e4ca68b7beda6d5 … 00a/irl3803.pdf) do you think it would be suitable?

No Idea. not enough experience or knowledge of all the little intricacies fets have to offer.
if you don't find that one, I guess you could also go for this one : IRLZ44PBF https://eu.mouser.com/datasheet/2/427/sihlz44-1768805.pdf
there's probably more compatible ones to be found depending on which supplier you consult.

Thank you for the feedback, let's wait if someone has more experience about the mosfet values.
Also I just got another weird warning, I finally was able to complete a run of winstone 99, it showed some weird warnings:

Also I noticed that the performance are not fair, sometimes quake2 gives me 50 fps, other times 30. It's too much for being a standard deviation.

Nemo1985 wrote on 2021-04-21, 15:27:

Thank you for the feedback, let's wait if someone has more experience about the mosfet values. Also I just got another weird war […]
Show full quote

Thank you for the feedback, let's wait if someone has more experience about the mosfet values.
Also I just got another weird warning, I finally was able to complete a run of winstone 99, it showed some weird warnings:
photo_2021-04-21_17-21-57.jpg

Also I noticed that the performance are not fair, sometimes quake2 gives me 50 fps, other times 30. It's too much for being a standard deviation.

fyi, EEVBlog, recently did a video about choosing a replacement FET

weedeewee wrote on 2021-04-21, 15:30:

fyi, EEVBlog, recently did a video about choosing a replacement FET

That's a precious information, I'm going to watch it.

In the meanwhile I have bad news, I disabled the l2 cache, changed the multiplier of the cpu (underclocked from 250 to 200), then fired the system again.
Well winstone crashed.
There is something very wrong with this cursed motherboard.
I also measured the mosfets temperature with a thermometer laser gun, they are at 25°C

Given it was working for a few months, this may be a different problem to the original one. If you haven't already, might be worth going back and checking with the other CPU you mentioned in the first post, and checking with memtest again. Also, those Winstone messages look pretty close to it saying the CPU changed from 200MHz to 166, to 133. Weird. Could it be an issue with the PLL, or the inputs to it?

Nemo1985 wrote on 2020-08-16, 20:16:

Thank you for your precious advice. I agree it is a very good board mainly thanks to the wide voltage selection. I will test th […]
Show full quote

Thank you for your precious advice. I agree it is a very good board mainly thanks to the wide voltage selection.
I will test the mb again before order the capacitors, as a matter of fact, I was able to overclock my k6-3+ to 600 mhz, but I just tested it with 1mb super pi without issues, while with higher voltage cpus (IDT) I had many issues.
Did you recap the whole board or just the bigger capacitors?
Do you have a link with the capacitors you used?
Thank you

Sorry, haven't visited the forum in a while.

I don't have a lint to the caps since I bought them from a local electronics supplies shop. I recapped the whole board, not just the large caps. I also added a couple of caps in unpopulated areas since my board did not have them but review samples of the board I saw online did.

This motherboard doesn't like the Rise MP6 cpu, it is the one that gives me more troubles (I know for sure it is compatible, since there is another fellow vogonians who built a rig with this cpu and motherboard).
I'm doing a roundup and so far: Rise MP6 the most problematic, Pentium MMX tillamook no issues, IBM 6x86MX (it's a cyrix MII) had issues with winstone and some slowdown.
I redid the winstone test with MP6 and it crashed twice (once again) with crashing of the program (not the OS), now I disabled the cache and it completed successfully twice without frequency errors according to Winstone.

Nemo1985 wrote on 2021-04-21, 15:27:

Thank you for the feedback, let's wait if someone has more experience about the mosfet values. Also I just got another weird war […]
Show full quote

Thank you for the feedback, let's wait if someone has more experience about the mosfet values.
Also I just got another weird warning, I finally was able to complete a run of winstone 99, it showed some weird warnings:
photo_2021-04-21_17-21-57.jpg

Also I noticed that the performance are not fair, sometimes quake2 gives me 50 fps, other times 30. It's too much for being a standard deviation.

Maby a clock/crystal problem?
At some point benchmarks wised up when it came to cheating with the time and changing crystals.
Maby thats what the error mess suggests?

Nemo1985 wrote on 2021-04-21, 18:50:

This motherboard doesn't like the Rise MP6 cpu, it is the one that gives me more troubles (I know for sure it is compatible, since there is another fellow vogonians who built a rig with this cpu and motherboard).
I'm doing a roundup and so far: Rise MP6 the most problematic, Pentium MMX tillamook no issues, IBM 6x86MX (it's a cyrix MII) had issues with winstone and some slowdown.
I redid the winstone test with MP6 and it crashed twice (once again) with crashing of the program (not the OS), now I disabled the cache and it completed successfully twice without frequency errors according to Winstone.

Retest the mp6 with cache enabled at a lower FSB, like 66 MHZ
It kinda seems that way though I doubt the cpu scales back the clock due to overheating.

weedeewee wrote on 2021-04-21, 19:06:

Retest the mp6 with cache enabled at a lower FSB, like 66 MHZ
It kinda seems that way though I doubt the cpu scales back the clock due to overheating.

The Mp6 won't overheat, it's the low voltage version, I have a beefy heatsink+fan, the cpu isn't even warm, anyway those stuff are too old to termal throttle.
Now i'm testing the cyrix without the l2 cache, but it crashes no matter what, I had enough for today.
tomorrow I will test with 66 mhz fsb as you adviced.

I had to restore windows from scratch, when winstone crash it always do it in a catastrophic way.
Actually testing the Cyrix\IBM v2.9 (2x100) cpu with l2 cache disabled, I swapped the ram from slot 3 to slot 1 (the ram bank is proved to be fine, also before it was working fine on slot 3):

Completed without errors but the frequency switching is still there.

L2 cache enabled:

Completed without errors but the frequency switching is still there.

Rise mp6 (2,5x100), l2 cache enabled: first time it crashed upon starting the program, second time, I had several applications crash:

So, in the end I suppose we have 2 different issues there.
The motherboard goes nuts with frequency and there are l2 cache stability issues but they are intermittent.

I'm considering that there could be a socket issue there, as Tiido supposed when he tried to fix the motherboard, the next step is find a guide about how to open it without destroying the plastic cover. I found a guide some time ago about opening a socket 370 but I'm unable to find it and when I tried to take off the plastic cover from a sloket it just tear down in pieces.

I'm very disappointed and low morale, I had the chance to compare all the ss7 cpus (well almost) but with so many troubles, I will need to postpone until I will find a working aopen motherboard

So I changed the crystal as adviced, once again it did not solve anything.
I ran memtest for like 6 hours without issues, I'm running quake in loop without issues, but as soon I try to use windows I get issues.
With a CF card i'm able to install windows fine but after a minute on desktop it just reboots itself.
I tried to install windows in a hard drive and it crash during the installation (once it just exited, the second mentioned a gdi error).

Using a pci video card, made stop windows to randomly reboot, but the frequency switching is still here.

Can anyone help me please?

If it works fine with a K6-3 or 3+ and the L3 cache disabled, then you need to replace those two cache SRAMs.

ph4nt0m wrote on 2021-06-01, 17:04:

If it works fine with a K6-3 or 3+ and the L3 cache disabled, then you need to replace those two cache SRAMs.

Actually the problem is with frequency switching and random reebots with an agp video card installed.
Probably the culprit is the pll chip.
The main problem is to find another one, since utsource sold me fake things and i'm still fighting with them to get my money back.
The original part is a Cypress W48S87-27AH, but I hope there are other compatible models.

To summarize:
With Cyrix\IBM M2 (V2.9) cpus there is frequency jumping
K6-3+ (V1.6) no frequency jumping
Cyrix 6x86L (V2.😎 no frequency jumping but the pc is almost every time unable to boot, the speak ring like a bell many short beeps
Idt Winchip (V3.52) suddenly crashed with the speaker doing many short beeps on boot.

PLL malfunctions are very occasional in my experience. If you borrow a high speed scope, it should help. I use a 200MHz Fluke 199C for such purposes, but this thing is very expensive. I suggest to double check all SMD components around the PLL because they are more likely to fail.